Characterizing artificial intelligence (AI) psychosis in a large academic medical setting: evidence of the new clinical phenomenon and the vulnerability of those in early phases of psychosis

Key Findings

• — Of 73 psychosis patients whose records mentioned AI, 28 were rated as experiencing AI psychosis; ChatGPT was the matching keyword in 53.6% of those cases
• — Most AI-psychosis cases were documented after the May 2024 release of GPT-4o; 60.7% of the AI-psychosis group were in a first psychotic episode, significantly more than comparison groups
• — "Amplifier" was the most common interaction rating (64.3%): AI most often exacerbated an existing condition by reinforcing distorted ideas rather than causing illness independently

Methodology Notes

EHR keyword search (e.g. "ChatGPT", "AI") across Vanderbilt University Medical Center records from 2022-12-01 to 2026-04-01; records excluded if not AI-related or if primary diagnosis did not include psychosis; three raters read clinical notes and applied four a-priori interaction categories. Preprint v1 posted 2026-06-08, marked published-ahead-of-print. medrxiv.org blocks automated fetchers; title, authors, date, and abstract verified via the official medRxiv API.

Sources